- AI chatbots appear to be providing less diverse responses, research claims
- Displaying the same answers has been likened to globalisation, where the same products and brands are available everywhere
- Such a move may increase convenience, but also reduce diversity
Could we be heading toward a real‑world Idiocracy? Repeatedly relying on AI for answers—often the same responses with identical viewpoints—may be narrowing human knowledge, according to a new study.
The research, which evaluated 27 different LLMs, found that even the top‑performing model delivered answers roughly 19 % less diverse than those returned by Google’s standard search. As a result, the information stream is less varied, offering a narrower range of answers, perspectives, and interpretations.
Researchers warn that this tendency could push certain information to the fringes of accepted knowledge while elevating other content as dominant. The ultimate outcome might be a society where people are less informed, more prone to simplistic thinking, and possibly even less intelligent than before.
Topic testing
The study was carried out by researchers at the Department of Computer Science (DIKU) of the University of Copenhagen.
To gauge how LLMs generate answers, the team examined 155 topics spanning everything from nuclear weapons to K‑pop. They employed 200 prompt variations derived from studies of real‑world chatbot use, producing about 70 million model responses.
Even the most varied output from GPT‑5 still fell 18.5 % short of the diversity seen in Google Search results.
“People are largely exposed to the same information over and over,” notes lead author Dustin Wright. “AI chatbots are therefore reshaping not only how we seek knowledge, but also which knowledge we can access.”
An Idiocracy?
Whether AI models are used within Google Search, DuckDuckGo, Bing, or directly queried via Gemini, ChatGPT, or other chat‑accessible LLMs, the research indicates a shift in how information is surfaced.
“We risk limiting people’s exposure to diverse viewpoints and a narrow knowledge base. This could foster a vicious cycle where popular content grows even more dominant while alternative perspectives fade,” says DIKU Professor and senior author Isabelle Augenstein.
She likens the effect to the downsides of globalization: “Today you can buy the same products and encounter the same coffee chains almost anywhere on the planet. That brings convenience, but it also erodes variety.”
Regarding the prospect that AI might shrink humanity’s collective knowledge and leave society in the hands of those ill‑suited to steer it, Professor Augenstein remains optimistic. She observes that a “knowledge collapse has not yet occurred,” adding that newer models tend to yield slightly more varied answers than their predecessors.
